AP Statistics: Central Limit Theorem

The Central Limit Theorem (CLT) is the statistical engine that allows us to make powerful inferences about populations using data from samples. It explains why the normal distribution appears so frequently in nature and data analysis, even when the underlying population data is not normal at all. Mastering the CLT is essential because it is the theoretical foundation for constructing confidence intervals and performing hypothesis tests—the core tools of statistical inference you will use on the AP exam and in real-world engineering and scientific applications.

The Core Idea: From Population to Sampling Distribution

To understand the CLT, you must first distinguish between three key distributions. The population distribution is the distribution of values for the entire group you’re interested in; it has a mean, often denoted by the Greek letter $\mu$, and a standard deviation, $\sigma$. This distribution can have any shape: highly skewed, uniform, bimodal, or even entirely irregular. When you take a single random sample of size $n$ from this population, you get a collection of data points. The mean of that sample is called the sample mean, denoted $\bar{x}$.

Now, imagine repeating this process an infinite number of times: take a sample of size $n$, calculate $\bar{x}$, and plot it on a new graph. The distribution of all these possible sample means is called the sampling distribution of the sample mean. The Central Limit Theorem makes a profound and specific claim about the shape of this sampling distribution.

Formal Statement of the Central Limit Theorem

The Central Limit Theorem states that for a population with mean $\mu$ and standard deviation $\sigma$, the sampling distribution of the sample mean $\bar{x}$ for random samples of size $n$ will have the following characteristics, provided that the samples are independent:

1. Mean: The mean of the sampling distribution is equal to the population mean. That is, ${\mu }_{\bar{x}}=\mu$.
2. Spread: The standard deviation of the sampling distribution, called the standard error (SE), is equal to the population standard deviation divided by the square root of the sample size. The formula is ${\sigma }_{\bar{x}}=\frac{\sigma }{\sqrt{n}}$.
3. Shape: As the sample size $n$ increases, the shape of the sampling distribution becomes approximately normal. This approximation improves as $n$ gets larger.

The revolutionary part is the third condition about shape. It means that no matter what the shape of the original population distribution—whether it’s the time until a machine part fails (right-skewed) or the uniform distribution of random numbers—the distribution of sample means will tend toward a bell curve if your sample size is sufficiently large.

Conditions for the CLT

The CLT doesn’t apply magically; specific conditions must be met. For the sampling distribution of $\bar{x}$ to be approximately normal:

• Random: The data must come from a random sample or a randomized experiment.
• Independent: Individual observations must be independent. In practice, this is often ensured by sampling without replacement. When sampling without replacement from a finite population, we check the 10% condition: the sample size $n$ should be no more than 10% of the population size $N$ ($n\le 0.10N$).
• Large Sample/ Normal Population: The required sample size for the CLT to "kick in" depends on the population's shape.
• If the population distribution is normal, the sampling distribution of $\bar{x}$ is normal for any sample size $n$.
• If the population distribution is not normal, the sampling distribution of $\bar{x}$ becomes approximately normal as the sample size $n$ increases. There is no universal threshold, but $n\ge 30$ is a common rule of thumb for the approximation to be reasonable, even for moderately skewed populations. For populations that are heavily skewed or have outliers, a larger sample size (e.g., $n\ge 40$ or more) may be necessary.

Applying the CLT: Computing Probabilities

This is where the CLT becomes practical. Once we know the sampling distribution of $\bar{x}$ is approximately normal with mean $\mu$ and standard error $\frac{\sigma }{\sqrt{n}}$, we can use z-scores and the standard normal distribution to compute probabilities.

The standardized value (z-score) for a sample mean is calculated as: $$z=\frac{\bar{x}-\mu }{\sigma /\sqrt{n}}$$

Example Problem: An engineer knows the output voltage of a circuit board has a mean $\mu =12.4$ volts and standard deviation $\sigma =0.8$ volts. The population distribution of voltages is slightly right-skewed. If the engineer takes a random sample of $n=40$ boards, what is the probability that the sample mean voltage is less than 12.2 volts?

Step-by-Step Solution:

1. Check Conditions: We have a random sample. The 10% condition is reasonable if more than 400 boards exist. The population is non-normal but the sample size ($n=40$) is large ($\ge 30$), so the CLT applies. The sampling distribution of $\bar{x}$ is approximately normal.
2. Define the Sampling Distribution:

• Mean: ${\mu }_{\bar{x}}=\mu =12.4$
• Standard Error: ${\sigma }_{\bar{x}}=\frac{\sigma }{\sqrt{n}}=\frac{0.8}{\sqrt{40}}\approx 0.1265$

1. Calculate the z-score: $z=\frac{12.2-12.4}{0.1265}\approx -1.58$
2. Find the Probability: $P(\bar{x}<12.2)=P(z<-1.58)$. Using a z-table or calculator, this probability is approximately 0.0571.
3. Conclusion: There is about a 5.7% chance of getting a sample mean voltage below 12.2 volts from a sample of 40 boards.

Common Pitfalls

1. Applying the CLT to the Sample Data, Not the Sample Mean: The CLT describes the distribution of the statistic ($\bar{x}$), not the distribution of the original sample data. A common mistake is to assume your single sample of $n=30$ data points will look normal. It might not. The CLT says that if you took many such samples, the collection of their means would form a normal distribution.
2. Forgetting the Conditions, Especially Independence: The $n\ge 30$ rule is useful, but it’s not a substitute for checking randomness and independence. If you sample 30 friends from your class without considering the 10% condition, your results may not be generalizable, and the CLT's conclusions may be invalid.
3. Misusing the Standard Deviation Formula: A critical error is using the population standard deviation $\sigma$ in the z-score formula instead of the standard error $\frac{\sigma }{\sqrt{n}}$. This drastically overstates the spread of the sampling distribution and leads to incorrect probabilities. Always remember to divide by $\sqrt{n}$.
4. Assuming $n\ge 30$ is Always Sufficient: While a good guideline, $n\ge 30$ may not be enough for populations with extreme skew or outliers. For example, the sampling distribution of the mean for a heavily skewed income population might still show some skew even with $n=30$. Always consider the context of the population shape.

Summary

• The Central Limit Theorem (CLT) states that the sampling distribution of the sample mean $\bar{x}$ becomes approximately normal as the sample size $n$ increases, regardless of the population's shape, provided the samples are random and independent.
• The mean of this sampling distribution equals the population mean (${\mu }_{\bar{x}}=\mu$), and its standard deviation is the standard error: ${\sigma }_{\bar{x}}=\frac{\sigma }{\sqrt{n}}$.
• The common rule of thumb $n\ge 30$ is used to justify the normal approximation when the population distribution is not normal, but it is not a guarantee—always assess skewness and check the underlying conditions.
• This theorem enables all statistical inference about a population mean. By standardizing $\bar{x}$ to a z-score $z=\frac{\bar{x}-\mu }{\sigma /\sqrt{n}}$, you can use the normal distribution to calculate probabilities about sample means, which is foundational for building confidence intervals and conducting hypothesis tests.